Acute myeloid leukemia (AML) is the second most common form of childhood leukemia and has the worst prognosis of all major childhood cancers. The nucleoside analog, Cytarabine (ara-C), is the most effective and widely used chemotherapeutic agent used to treat AML. However, wide inter-patient variation in its treatment response, development of resistance, and severe toxicity remain major hurdles to effective ara-C chemotherapy. Inter-patient variation in expression and/or activity of ara-C pharmacokinetic (PK) and pharmacodynamic (PD) genes is likely to contribute to the variability observed in ara-C treatment outcomes. During the last funding period of this application, we evaluated 12 PK genes in ara-C metabolic pathway and identified genetic polymorphisms that could explain a substantial portion of the variability in ara-C clinical response and could be of similar prognosti relevance as currently used risk stratification factors in AML patients. We found that genetic variation in PK genes may explain 11% of variation in prognosis (event-free survival, EFS) after accounting for 6 well established prognostic factors that cumulatively explain 18% of variation in EFS. To more fully understand the mechanisms contributing to variability in ara-C treatment outcomes, we now propose to study genes involved in ara-C pharmacodynamic response, which unfortunately have not been well studied so far. We have developed a novel statistical method, PROMISE (PRojection Onto the Most Interesting Statistical Evidence), which dramatically increases our power to make important pharmacogenomic discoveries by identifying genetic features with a biologically meaningful pattern of associations with multiple pharmacologic and clinical endpoints. As a first step, we will use PROMISE to identify diagnostic leukemic blast gene expression signatures associated with in vitro leukemic blast sensitivity to ara-C as well as multiple clinical outcomes in AML patients (Aim 1). Following functional validation of candidate ara-C PD genes, we will identify and validate the clinical and prognostic significance of polymorphisms in ara-C PD genes in AML patients from multiple independent cohorts (Aim 2). Since epigenetic mechanisms such as DNA methylation have been suggested to influence gene expression in AML, we shall also evaluate DNA methylation for its role in regulating expression of ara-C PK and PD genes, and in influencing clinical outcomes in AML patients treated with ara-C (Aim 3). Finally, we will identify, and validate in independent cohorts of AML patients, the pharmacogenetic and pharmacoepigenetic markers that supplement currently known prognostic factors in an integrated system of risk assignment for purposes of determining treatment intensity. Understanding the interplay of genetic and epigenetic factors in mediating ara-C response and their integration into current prognostic features would present an opportunity to increase our accuracy in forecasting therapeutic outcomes in AML and allow more tailored, risk-stratified treatment approaches - a major advancement over current strategy.